摘要 |
PROBLEM TO BE SOLVED: To make processable the voice recognition and voice synthesis in common, by providing a voice waveform node for the input/output of the voice waveform and a word symbol node for the input/output of a character line, and directly inputting the voice waveform to a neural circuit. SOLUTION: The voice synthesizer is constituted of a voice waveform node 1, a word symbol node 2, and intermediate nodes 3-6. The voice waveform node 1 receives the input voice as the voice waveform and outputs it to individual nodes in voice recognition and outputs the wavefrom of the synthesized voice according to the outputs from the individual nodes in voice synthesis. The word symbol node 2 outputs the recognition result of the inputted voice waveform according to the outputs from individual nodes in voice recognition and receives the outputted synthesized voice in a character line and outputs it to individual nodes in voice synthesis. The intermediate node 3 converts and defines the necessary process and data as the function of nodes and for weighting between the nodes. The voice waveform is directly inputted to a neural circuit, and the whole processing of voice recognition and voice synthesis is done in the common neural circuit. |